Power Supply from ATX, DIY

Well there's a few tutorials and such available on-line on how to turn an ATX power supply into a laboratory power supply. I don't know if this thread is meant to be 'blog' like or not, because I'm more than likely going to need help/advice along the way.

So I went on craigs-list and found an old dell power supply: http://www.atxpowersupplies.com/160-watt-dell-power-supply-p...


It' says it's ATX and lists it's out-put's. I know nothing about electronics so this will be an interesting adventure. First things first I'm going to try and find a wiring schematic to figure out what wires are what voltage.

If anyone's done this some advice would be helpful, no your not responsible if I zap myself to death, I am .

* Update * Borrowed from - http://www.ehow.com/how_2251617_add-variable-voltage-atx-bas...

On the ATX connector there will be a green wire. That shorted to any of the grounds will turn the unit on. Pop the thing open and see if all the rails are one. Sometimes even the cheap power supplies have more than one independent 12V rail inside. Note some supplies need a parasitic or ghost drain in addition to the power on wire to start. Connect an old drive or 12 volt device to the unit if it doesn't start with just jumpering.

Upon closer inspection it has an independent line for processor power. It will likely have more than one rail. It was almost standardized to keep drives powering up from lowering the voltage going to the processor.

Open it up and take a picture of the pcb where the wires connect. I can guide you very well in all things computer

@woelen - Wow woelen you rock, thanks for the link and taking the time to do that. I'm going to keep the 12V even though I'm not sure exactly what kind of experiments I'll be using it for. I just found this neat e-book called "Handbook of Electrochemistry", and it really seems like a useful mind-set/skill to be in as a home scientist.

A lot of interesting reactions can be done with little to no reagents(saves cost), and it seems like there are a lot of experiments and things to still be found out in the field.

@Regolith, the PCB is pretty hard to see, there's all kinds of epoxy(?) and such covering connections. And for some reason my camera isn't recognizing this SD card . Sounds like another project to work on for today hehe.

Anyhow,
I'm picking up some binding posts tomorrow, a little LED /w resistor, and hopefully some resisters(depending on cost only have $5 to my name hahaha). I'll update this as things progress along surely with questions, ideally with instant success.

Nice Woelen! thats hardcore it's almost like you have threads here with entires about chlorate cells. How crazy would that be ? Just a side note from my own work with computers. If your duplicating these results you actually don't want the craziest highend supply you can get.

I've seen far more regulation circuitry of late in the higher end 3 and 4 (12 volt) rail supplies that didn't allow me to use it for electrolysis. I was attempting a torch based on oxyhydrogen (got it working eventually with a rewound MOT, inelegant but functional) and the circuit would cut out after several minutes of water cracking. I needed to bypass several built in regulators that while it wasn't overheating, was assuming the system was in malfunction. Also the 12 volt rails (in this model) weren't capable of stacking amperage due to the way the transformer had been tapped inside(and instantly shutdown if cross leakage was detected). I haven't yet gotten around to rewiring the transformer, which is doable but unnessessary work.

Stick to the cheaper models with less bells and whistles for these and use Woelens guide (I seriously repect those who make guides like that for the uninformed) for the wiring.

That mention of the cross leakage and it shutting down was interesting Regolith.

I've been trying to power up some 12V circuits recently that could really do with tens of easily adjustable amps. I have three power supplies already, but the load lines of them keep getting in the way as the load gets even half way towards the supply's maximum wattage. I've seen countless pages about using them as is, or sticking a linear regulator on the output, but I was curious to know how close one of these could get to approximating one of those multithousand pound rack mounts people like Lambda sell. Someone has already done most of the hard work in getting the dirty AC into DC, roughly stabilising it, adding some protection, filters and so on.

I began reading around and discovered, in the wiki titled power supply (computers), some useful information.

1. Multiple 12V power supply rails are separately current limited as a safety feature; they are not generated separately. Despite widespread belief to the contrary, this separation has no effect on mutual interference between supply rails;
   
   
2. The ATX12V 2.x and EPS12V power supply standards defer to the IEC 60950 standard, which requires that no more than 240 volt-amps be present between any two accessible points. Thus, each wire must be current-limited to no more than 20 A; typical supplies guarantee 18 A without triggering the current limit. Power supplies capable of delivering more than 18 A at 12 V connect wires in groups to two or more current sensors which will shut down the supply if excess current flows. Unlike a fuse or circuit breaker, these limits reset as soon as the overload is removed.
   
   
3. Because of the above standards, almost all high-power supplies claim to implement separate rails, however this claim is often false; many omit the necessary current-limit circuitry,[5] both for cost reasons and because it is an irritation to customers.[1] (The lack is sometimes advertised as a feature under names like "rail fusion" or "current sharing".)
   

One simple solution to regulation is to fight back, with a jumper! Important to remember that rebellion usually involves a building on fire.

But this is all just groundwork for my far more epic scheming.

Having a look inside, I discovered two IC's. One a voltage comparator and the other a PWM switch mode controller; as one might expect in a SMPS.

I began chatting with some nerds from the darkside of electronics about the possibility of hacking the switch mode section it's self, as opposed to sticking big resistive dividers or linear regulators (yeah.... resistive dividers ) on the output and needing huge heatsinks.

Hey presto! Someone has been daring enough to start fiddling with the feedback for the controller.

By playing with the voltage the feedback pin is getting, you can not only adjust the rail voltage from 12 down, you can run it up to about 50V, just before the output caps explode and blow hot goo towards you.

That's exciting! Because it conserves the space and weight saving potential of the SMPS, but provides you with an adjustable rail voltage.

One initial problem is, the controller will think the rails have gone over their limits, and trigger the shutdown. Solution? You know it. Play with / disable the over voltage on the rails.

This is starting to show some really cool potential. You now have the rails adjustable for a constant volage, and (provided you don't entirely disable it, but alter it instead), the O.V.L. that the rack mount supplies for labs feature.

These newer models have the current limit as well. So I expect that could also be butchered into being adjustable, or even toggled back to the controller to make it a constant current source? Or, if you double up the controller or swap it out, both?

I am unsure what playing with the pulse widths to this extent does to the noise spectrum and such of the power supply. It is fix frequency, so I doubt it'll become horribly inefficient provided you're drawing the normal currents through the magnetics (e.g. they won't saturate). The filters will be tuned to the frequency it's running at, so hopefully they shouldn't be too badly skewed. And, for a lot of chemistry work anyway, noise isn't really an issue. A bigger problem may be it dropping out of regulation at low voltages / currents. Soltuion? Use a different rail or just a separate linear supply will do if the power requirement is low. I also don't know precisely how much voltage droop the rails show when they're at full load. But, based on them being able to reach at least 50V with the PWM all the way up and it unloaded and that the silicon they're usually powering needs a stable supply at varying (sometimes full) loads, I would expect it's not a whole lot.

I was also thinking, whilst adding a huge linear regulator to the output to adjust the bulk of the voltage or current is defeating the point of using a SMPS in the first place, we could have one tracking the output, dropping only it's own voltage drop to stay in regulation. That would not be to adjust the output, but it could be useful if the output is a mess, to clean up the noise and make it even more useful for electronics work. As a tracking linear stage on the output would only need enough voltage drop to keep it's self in regulation, it shouldn't require huge heat sinks.

If you wanted to get brutal on it, the super regulator.

This is why I was interested in what you'd got up to with them. As there is now a lot of stuff packed into there that could supply a number of the futures in the high end rack mount things.

There's a question of when modifying something becomes more complex or expensive than starting from scratch. But I don't think many of the things I've mentioned would be comparable to building such a compact SMPS from scratch. Things like altering the rail voltage essentially involve finding the correct point and dropping a pot across them. Unlike finding specific magnetics and tuning a SMPS.

If the rails from multiple supplies could all be stacked up to the limit of a normal house socket, that'd be a serious bit of kit! A comparable Lambda Genesys has got to be a few thousand dollars.

Beginning to wish I had my 123 to hand now.

The target


The starting point


This could really use an injection of Tim "MegaWattZ" 12AX7, as he's built a 10kW induction heater, from scratch, he's probably the most well versed in the black art round these parts.

[Edited on 28-2-2011 by peach]

Cool, excellent primer on a home-built PS, Woelen! I have a bunch of old power supplies in the shed and I know my next weekend is booked making one of those according to your thoughtful instructions.

I know that there are fewer and fewer of those available, but back in the the days of 386, 486 and early Pentium1 processors, the standard AT (not ATX) power supplies were wonderful to work with (the ones with the two flat P5/P6 connectors and the hardwired 110v pushbutton power switch).

They had less protection circuitry and many of them could be operated with little or no load. I just added a plain 12v light bulb connected on the 12v bus , which acted as a minimal load and a power-on pilot light and that was it. I still use this supply often, I built it over 20 years ago! However, I don't want to use it for electrolysis, so I'll make a new one just for that purpose.

So if you see people throwing away those old obsolete computers, grab' em! they're chock full of neat goodies, from powerful fans, to gold-plated processors to useful screws and hardware.

Usually, when I throw away a computer, it's a cannibalized empty shell with not much else than the case and the stripped-off boards!

Robert

Yep.

The output capacitors are not going to do well running them at high voltages. Picking any random supply out and just turning the rails up, I expect quite a few people would report blown capacitors.

I can count 19 electro caps on the board in front of me. Two of these are large 200V caps, which I suspect are the mains side. There are a number of small ones dotted around between the two heatsinks and near the two IC's. These are rated at 50V and are likely to do with the controller and switching side. At the right of the board is another large cluster. The majority of these are rated at 10 or 16V, and they'll be the filters for the output, which will be exposed to the higher rail voltages. Electro caps are usually derated by 1/3 to prolong their life and avoid bursting them with over voltages. So, with this example, there is basically no room for going over the limit without replacing those with higher voltage values.

Producing high voltages, such as 50V, from the supplies was not primarily of major interest. As the output of these is floating, I suppose multiple supplies could be carefully stacked to increase the voltage. And the winding on the transformers appears to be one continuous deal, with taps for each rail voltage.

In terms of electrochemistry, the lower voltage rails (like +5) would be a much easier place to start given that they're closer to the value the supply is supposed to run at (reducing the chance that it'll drop out of regulation as you turn it down) and they're usually rated for a higher current, which is what you'd be after for electrolysis, plating and so on.

Does anyone have experience stacking the outputs from more than one supply? Here are some photos and my quick guesses at the SMPS for those nerds into their electronics and curious about the potential.

This is a supply from 2003



The mains is through the black and white wires under my thumb. They immediately hit a fuse and, something lacking in my 30 amp industrial supplies, an NTC, to limit the surge when it's switched on. Nice.

In the bottom right there is a small blue cap rated at 2kV, which will be a line filter, and the AC then goes to the bank of four diodes just to the left of it. The two big caps I am guessing are also to do with smoothing the rectified DC. They must be in series with each other right? Because they're rated at 200WV, and my initial guess is they're in parallel with a bridge rectifier, which is being fed 240Vac, which will give more than that in DC. Alternatively, the rectifier is a half bridge.

The tiny resistor to the right of the first big cap may be a bleed. The tiny ceramic beside it may also be for diode noise or bypassing. There is another, considerably larger, ceramic just behind the two big electros to the left. Probably also part of the line filter.


This is where things start getting really impressive if you're a nerd. The three black bits of silicon on the heatsink to the left are the switching transistors, which begin chopping the DC up at high speed and feeding it through.

The big transformer at the back is almost certainly the power transformer with the secondary taps that feed each of the rails at the end of everything.

The two smaller transformers I'm thinking are the isolation / feedbacks for the controller.

I'm not entirely sure what the thick wire, few winds inductor is (without consulting the schematic). It could be to filter noise, pulse shaping or to help stability in some way. As there's no other winding on it, it won't be for feedback.

All of the caps and diodes in this stage are likely related to feedback and switching.

That power transformer is actually rather small indeed given what is flowing through it. The beauty of switch mode.


The area at the bottom right, surrounded by resistors and tiny green caps is the controller it's self.

Only two IC's, with the one on the left being the KA7500B mentioned in the page I linked to (the switch mode controller from fairchild). The one on the left is the comparator.

This is the datasheet for the KA7500B controller. In this photo, it's upside down. The two pins needed to squiffy the rail voltages are the second from the right on the top and the third from the left on the bottom.

If I could find a roughly 10k pot, I'd be tempted to try it with this busted up board, since one of the rails will still be functioning.


The output section.

The AC from the mains has now been rectified to DC pulses, smoothed by the input capacitors, chopped up into variable width high frequency pulses by the first transistors and controller, fed through a step down transformer as AC at a new frequency and arrives..... here.

The second heatsink, on the left, is lined with rectifiers to convert it back to DC. The toroidal inductors form an LC filter, which is tuned to the frequency of the chopping to make it as small as possible. The remaining capacitors are effectively the voltage limit of the rails, as they're very close to the normal rail voltages in this example and would vent (pop / explode / release the magic purple smoke), through those crosses on their tops if they were driven into an over voltage state. There are quite a few of them, one for each rail.



[Edited on 28-2-2011 by peach]

@woelen - Thanks for the advice again. So far I've invested $7 and a cord should cost $1.99 + shipping. So we'll see. If it doesn't work the idea behind the design was actually based on this hehe. If it doesn't work I can still get a different power supply cut the cables and resolder them to the cables already wired in the external box.

@regolith, thanks for the clarification on that . Soon as I get the power cord and can borrow my friends voltmeter I'll update this thread. Soldering to those binding posts seriously sucked as far as soldering goes but it all worked out .

Borrowed the cord from my parents computer. I plugged it in and there was nothing happening when I flicked the switch. My heart sank. Then I realized it was on 230 and not 115. Plugged in and flicked on the SPST switch the LED came on, the fan on the PSU spun. Good news!

tested a red LED(no resister) on the 5v and ground. Light came on.

tested the LED on the 12v and the light came on and quickly blew out hahaha.

So I'm assuming it's up to standard. I'm going to try to take it into school and see if they have a volt meter there I can test this sucker with. I didn't use the 10ohm 10watt resister or fuses. Just rigged it up as is.

[Edited on 4-3-2011 by smaerd]

I once used an old power supply for a chlorate cell, I didn't use any power moderation I just hooked up my electrodes on the 12v rail. About an hour into the electrolysis there was a massive bang and the safety switched turned off the power to all of the power points in my home. Blew a fuse. Valuable lesson learnt there.
Now I have a proper power supply where I can regulate the voltage and it makes Potassium chlorate quite well with a Carbon anode, it's a good idea to make lots of chloride solution and keep topping your cell up as it gets quite hot, the water evaporates and your anode corrodes a lot, the chloride is used up and there is a lot more resistance in the electrolyte which adds to the heat...

12v, 10A works great. from memory the ratio of potassium chloride to water was about 1:4.

As I said you are best to keep topping up your cell, once about 1/2 to 3/4 of the chloride has been converted the current slows down and the solution starts to boil, if you were to run your cell with refilling to keep it at 10A for 48 hours you would get heaps of potassium chlorate.

Oops went way off topic into chlorates there, sorry!

[Edited on 6-3-2011 by trezza]

@ woelen

Whoops! Actually, I just noticed there isn't a brown wire in my power supply! Here's the pinout I have:



So I guess there's no 3.3V sense wire in this supply. All other wires seem the same according to your instructions on your site. as I mentioned, it's an early 20-pin PS, I like it because the case is bigger than normal and so is the fan.

Robert




[Edited on 6-3-2011 by Arthur Dent]

Quote: Originally posted by Regolith

It sounds however like it's pretty stock still ? No modifications to the power sections, as in it's still wired for its factory voltages and regulators ?

Quote: Originally posted by Arthur Dent

@ woelen Whoops! Actually, I just noticed there isn't a brown wire in my power supply! Here's the pinout I have: So I guess there's no 3.3V sense wire in this supply. All other wires seem the same according to your instructions on your site. as I mentioned, it's an early 20-pin PS, I like it because the case is bigger than normal and so is the fan. Robert [Edited on 6-3-2011 by Arthur Dent]

Quote: Originally posted by smaerd

I'll just buy a volt meter one day when I can afford it lol. [Edited on 6-3-2011 by smaerd]

Figure I kind of owe it to you guys to at least upload a picture of the working apparatus(now that the camera is back). Works great just need to find/afford some better electrodes then pencil graphite(erodes way to quickly), etc. They eroded too quick before a test KClO3 electrolysis could complete, but the water does smell nice and mm mm chlorinated heh. One day I'll have the cash to experiment.



[Edited on 22-3-2011 by smaerd]